The use of Artificial Intelligence (AI) and Machine Learning (ML) in chemical exploration is nothing new. For decades people have sought to use informatic tools and algorithms to speed up discovery projects.
However, only relatively recently have we seen a significant surge in the use of these technologies in the Advanced Materials space, as processing speeds and ML models have improved over time. We refer to such application as Materials Informatics (MI).
MI can accelerate material discovery significantly: predicting suitable compounds or compositions for a purpose; optimizing formulations; and even predicting properties of given compounds or compositions.
Refining IP Strategy
Such rapid generation of ‘possibilities’ can open a lot of doors, but can also lead to information overload. There may be significant implications for IP strategy related to the project where MI is deployed.
What?
Firstly, there is the problem of ‘what to protect’. In many jurisdictions it is possible to obtain a patent covering a given AI-driven methodology, or even algorithmic implementations, which can give very broad and useful protection. That or those may be protected as well as the novel materials suggested and investigated.
However those materials might not necessarily all be of interest. Perhaps only a small subset has been validated ‘in real life’. Perhaps many are, in fact, already encompassed by other IP (and hence represent an FTO risk). Perhaps they are so disparate in structure that many patent applications would be needed to cover them all (which would be very costly).
Therefore, we must build some discernment into any filing programme.
When?
Secondly, we must consider the possibility of ‘when to protect’. Staging costs can be financially vital, and so it may not be desirable to file all patent applications at the same time. After assessing all the IP ‘available’ in the project, we must consider how it can be split and protected in stages without adverse effect to later filings.
In the meantime – or indeed, in some cases, in perpetuity – information may be kept as a trade secret. It may be that, for example, an AI model and methodology are kept secret, while only certain investigated materials are subject to a patent application. Then, a second generation of materials can be subject of a further application later, once experimental verification has been conducted.
A ’trade secrets’ focussed approach may be favoured by larger companies, for whom patents are not important for securing investment. Smaller companies may prefer instead to publish and protect their MI methodologies, to assist with attraction of investors and commercial opportunities.
Where patenting is not pursued, some form of defensive publication (to try to prevent competitors from patenting in a certain area) can be considered.
Where?
Thirdly, we must consider ‘where to protect’. It is clear that the US, Europe and China are the most popular jurisdictions for protection of MI technologies. However, there are significant presences in Japan, Korea, and elsewhere. Protection in all of these would be relatively costly, but could be important for securing a long term commercial future. This compounds the importance of strategy, as in the first and second points, discerning and planning timings of filings and disclosures.
Many of these considerations are in the background to any innovative endeavour. However, the volume and speed with which MI can provide potentially protectable IP massively increases their importance.
